For senior engineers and architects moving into AI
Learn production AI from a platform that is one.
The coach that tutors you is the same bounded agentic loop the curriculum teaches — built in public on Anthropic, AWS, and Cloudflare. Bring the years you already have; add the AI-native layer on top.
- Free sample — no signup
- Every claim cited
- Cancel anytime
A profession, not a certificate. Cert checkpoints — CCA-F · AIP-C01 · CAPA — prove it; they are not the point.
Retrieval question
Your agent returns different answers to the same prompt across runs. What actually fixes it?
Pick one — the coach won’t just say “wrong.”
You already ship production systems.
What you are missing is the AI-native layer: the agentic loop, evals, cost and routing, the trust boundary, deploying models across clouds. Most courses start from zero and waste your time. This one assumes the years you already have.
Who it's for
Where you start, where you land.
AI engineer / architect
Add the AI-native design layer on top of strong engineering.
AI platform engineer (AgentOps)
Run agents in production — observability, cost, reliability.
AI architect on Bedrock
Translate cloud architecture into GenAI architecture.
AI architect (fastest conversion)
Your systems thinking transfers almost directly.
The profession, not a skill
What an AI engineer/architect actually does.
These are the jobs you will be able to do — not a list of exam topics.
- Design an agentic system — the loop, orchestrator and sub-agents, tool boundaries — and justify the pattern.
- Design and run an eval suite — the single biggest "actually shipped" signal.
- Cost-model a multi-turn agent and build a model router (Opus→Sonnet→Haiku) with prompt caching.
- Threat-model against the OWASP LLM Top 10, especially LLM06 Excessive Agency.
- Deploy and operate on all three platforms — Bedrock/AgentCore, Claude + Agent SDK + MCP, Cloudflare edge.
- Scope AI regulatory obligations into a design — risk-tier a system across the EU AI Act and the US patchwork, and pin the retention, training-use and inference-region terms that decide whether a build is lawful.
How it sticks
Built so it stays in your head.
Mastery-based, not time-based
Advance on demonstrated competence — an 80% rolling accuracy gate fronts each module.
Spaced retrieval
An SM-2 scheduler resurfaces questions right before you would forget them.
A senior-calibrated AI coach
Haiku for hints, Sonnet for tutoring, Opus for mock interviews.
Provenance
Every question cites its source. We never invent exam content.
See the coach think
It names the misconception — not just "wrong."
Miss a question and the coach diagnoses the model behind the mistake, then replaces it. Here is how it answers a classic one.
Your agent returns different answers to the same prompt across runs. What fixes it?
Set temperature to 0 so the output is deterministic.
You're conflating determinism with correctness. Temperature controls sampling variance — turn it to 0 and you don't get the right answer, you get one fixed answer that can be reliably wrong. Reproducibility is a property of your eval harness and tool contracts, not a decoding knob. So: what would you measure to know the answer is correct, not merely stable?
Grounded in the lessons, cited, and graded never to hand you the answer key.
Why you can trust it
Engineered against the slop.
The web is filling with AI-generated courses that confidently teach the wrong thing. This one is built not to.
Every claim is cited
Each lesson and question carries its sources — primary docs and official exam guides. Nothing is invented.
The coach is eval-gated
Its answers are graded for grounding, for targeting the misconception, and for never leaking an answer key — on every prompt change.
A human approves every change
An agent drafts updates from the canonical sources as they shift; nothing reaches you until it is reviewed.
Early signal
From professionals making the jump.
Curriculum
One integrated path, not a bundle of courses.
56 modules live — building in public- Track 0Mindset Shift3 modules
- Track AThe Bridge5 modules
- Track BAgentic Systems — the core9 modules
- Track CAnthropic Deep Dive8 modules
- Track DAWS GenAI Architecture13 modules
- Track ECloudflare AI Platform7 modules
- Track FIntegrative Capstone & Interview5 modules
- Track GAI Regulatory & Compliance Architecture6 modules
- Total56 modules8 tracks
Checkpoints
The certs evidence the profession.
Certs are checkpoints, not the destination — the destination is the job.
CCA-F
Anthropic · Claude Certified
Pass 720 / 1000- D1 Agentic Architecture27%
- D2 Tool Design & MCP18%
- D3 Claude Code Config20%
- D4 Prompt Eng & Structured Output20%
- D5 Context & Reliability15%
AIP-C01
AWS Certified AI Practitioner
Pass 750 / 1000- D131%
- D226%
- D320%
- D412%
- D511%
CAPA
Cloudflare AI Platform · Academy credential
Practical capstoneNo official Cloudflare AI certification exists yet. CAPA is a synthesized Academy credential with a mandatory practical capstone — we say so honestly rather than imply an exam that does not exist.
From the founder
A senior making the exact jump you are.
I'm Wibo, an engineer in Amsterdam, moving into AI architecture the same way you will — by building production systems on Anthropic, AWS, and Cloudflare. This platform is that build, and the curriculum is the map I draw as I go: the patterns that hold up, and the potholes, marked before you hit them. The build is the curriculum.
— Wibo · Amsterdam, NL
Pricing
One subscription. The whole profession.
Start free with a real lesson. Pro adds the full path — with a 14-day free trial, cancel anytime.
A real lesson and the public guides — no card.
- A full sample module, start to finish
- Every public guide & the curriculum map
- Sources cited on every claim
$198 billed yearly · save 43% · cancel anytime
Every track, the AI coach, and everything that ships next.
- All 56 modules across 8 tracks
- Senior-calibrated AI coach — Haiku, Sonnet, Opus
- Spaced repetition + 80% mastery gates
- Cert prep — CCA-F · AIP-C01 · CAPA
- New tracks weekly — access moves with the field
Become the AI engineer your team is trying to hire.
A subscription, not a one-off — the field moves weekly, and your access moves with it as lessons, exam weights, and model guidance stay current.
Cancel anytime. No card to try a lesson. Sources cited on every claim.
Or get notified when new tracks ship.